Pre-trained CNN models as feature extractors for Video Emotion Recognition

What this is about ?

This repository contains the work related to the research paper titled Efficiency analysis of Pre-trained CNN models as feature extractors for Video Emotion Recognition.

Following are the names of Authors:

• Diksha Mehta
• Janhvi Joshi
• Abhishek Bisht
• Pankaj Badoni

Appropriate citations can be downloaded from the github page of this repository or alternatively from citations.cff.

How to use this work ?

• Two configuration files config.ini and fusion_training_hyparam.json are required under the utils directory for this project to run. The details of these files can be inferred from source code itself or by can be requested by sending a mail at abhishekbisht@gmail.com.
• config.ini is needed to provide the directory paths such as those of datasets.
• fusion_training_hyparam.json is needed to supply the aprropriate hyperparameters for training.

Acknowledgements

This work uses the following external resources.

• haarcascade_frontalface_alt2.xml taken from opencv.
• shape_predictor_5_face_landmarks.dat taken from the repository ageitgey/face_recognition_models by Adam Geitgey.
• Datasets
  • SAVEE
  • RAVDESS
• Several research papers were also referred and the authors have been appropriately credited in the research paper associated with this work.